INDEX

Explanations

words related to communication and discussions, possibly in a professional or formal context

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

Data Type

torch.float32

Hook Point

blocks.8.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

inel

-0.70

 Bridges

-0.68

 Dickinson

-0.65

toe

-0.65

gart

-0.64

gra

-0.64

sted

-0.63

onz

-0.59

mon

-0.59

ogun

-0.59

POSITIVE LOGITS

 happening

1.14

 transpired

1.03

 happened

0.94

 happ

0.92

 happen

0.92

 happens

0.83

 Happ

0.79

 wrong

0.74

 done

0.71

wrong

0.71

Activations Density 1.702%

words related to communication and discussions, possibly in a professional or formal context

No Comments

No Known Activations